# Setting Up Azure GPT-5 model

## Choosing the Model

## Available Models

After thorough benchmarking, Unique has taken the decision to only vet the main GPT-5 model, as the output quality of `mini` and `nano` models did not prove to be up to our required standards.

Hence you can set the language model to the GPT-5 (2025-08-07) Azure model using the following model identifier: `AZURE_GPT_5_2025_0807`.

For the complete list of approved Azure models, please refer to [Unique’s official list of supported models](https://docs.unique.ai/administrators/models-llm-management/how-we-evaluate-models/llm-availability-overview).

## Model Restrictions

The GPT-5 model only accepts a temperature value of `1` **when reasoning effort is enabled** (i.e., set to any value other than `none`).

> **Important**: The Unique toolkit **no longer automatically overwrites** the temperature setting. When reasoning effort is set to a value other than `none` (e.g., `minimal`, `low`, `medium`, or `high`), the **space admin must manually set the temperature to** `1` in the space configuration. Failing to do so may result in API errors.

### Reasoning

GPT-5 models offer 4 levels of [reasoning effort](https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/reasoning?tabs=gpt-5%2Cpython-secure%2Cpy#new-gpt-5-reasoning-features): `minimal`, `low`, `medium`, and `high`. Reasoning can also be disabled by setting the value to `none`.

For speed and performance purposes, **Unique has set the model reasoning effort to** `minimal`.

Please contact our Customer Success team to modify the reasoning effort settings.

> **Note on Code Execution Tool**: When the **code execution tool** is enabled on a space, the GPT language model is switched to the **Response API**. In this mode, the reasoning effort parameter name changes from `reasoning_effort` to `reasoning.effort`. Space admins should ensure the correct parameter name is used in space configurations when the code execution tool is active.

# Model Configuration in Spaces

## For Spaces 2.0

### Changing the Model to GPT-5 on Existing Spaces

Go to `Space Management` \> `Space` \> `Configuration` \> `Language Model`.

Replace the existing model with `AZURE_GPT_5_2025_0807`.

Click `Update` to save the changes.

### Creating New Spaces with GPT-5

For space creation with Spaces 2.0, you can refer to the [space creation documentation](https://docs.unique.ai/administrators/space-management/create-spaces).

Go to `Space Management` \> `Create Space` \> `Unique AI Chat` \> `Configuration` \> `Language Model`.

Set `Language Model` to `AZURE_GPT_5_2025_0807`.

Click `Create` to complete the space creation.

## Editing Custom Spaces

Go to `Space Management` \> `Custom Space` \> `AI Assistant` \> `Edit`.

Find the main language model configured in the json object, and replace the language model with `AZURE_GPT_5_2025_0807`.

**Important**:

- If the **reasoning effort** is set to any value other than `none` (e.g., `minimal`, `low`, `medium`, `high`), you **must set the temperature to** `1` on the Space 2.0 Configuration UI, or **remove temperature** in the JSON configuration on a Custom Space UI. The toolkit no longer overwrites this automatically.

- When the **code execution tool** is enabled, the model uses the **Response API**. In this case, the reasoning effort parameter in the additional LLM options must use the name `reasoning.effort` (with a dot) instead of `reasoning_effort` (with an underscore).